🎭 Browser Sentiment Analysis

Loading sentiment analysis model...

🔬 Technical Specifications

Model: DistilBERT Base Uncased

Parameters: 66M (distilled from 110M BERT-base)

Architecture: 6 layers, 12 attention heads, 768 hidden size

Vocabulary: 30,522 WordPiece tokens

Training Data: Stanford Sentiment Treebank (SST-2)

Accuracy: 91.3% on SST-2 test set

Model Size: ~250MB ONNX format

Inference: WebAssembly + WebGPU acceleration

Latency: ~50-200ms per inference (browser-dependent)

This prototype demonstrates client-side transformer inference using Transformers.js. DistilBERT retains 97% of BERT's performance while being 60% smaller and 60% faster.